Customer service teams report frequent errors, miscommunication, and inefficiencies, highlighting the need for human intervention in frequently complex customer engagement scenarios.

Despite the promise of generative AI (GenAI) technology to revolutionize call center operations, frontline staff are increasingly vocal about the bots’ current shortcomings, revealing a gap between management expectations and real-world performance.

A recent Cornell University study (peer-reviewed) of a major Chinese utility company call center, involving interviews with customer service representatives (CSRs), team leaders, and supervisors, paints a sobering picture.

Problems cited by CSR teams include three major areas:

  • Frequent intervention required: Workers had reported that AI assistants frequently misinterpret customer speech, especially when faced with diverse accents, rapid speech, or sequences of numbers. Said one respondent: “The AI assistant isn’t that smart in reality. It gives phone numbers in bits and pieces, so I have to manually enter them.” Homophones and nuanced language further trip up the system, often requiring human correction.
  • Erroneous emotion-recognition: Touted as a way for AI to assess caller sentiment, this function has also fallen short. GenAI routinely misclassifies normal speech as negative emotion, or relies too heavily on voice level as an indicator of attitude. As a result, many staff simply ignore the AI’s emotional tags, relying instead on their own judgment.
  • Inaccurate processing: While AI-generated call summaries and prefilled content were designed to reduce manual labor, CSR workers had found themselves spending extra time editing or deleting inaccurate information. “AI-generated outputs introduced structural inefficiencies in information processing because most AI-prefilled content required manual correction or deletion,” the study reported. This not only slows down workflows but also increases the learning burden on employees, who must adapt to new systems while maintaining accuracy and customer satisfaction.

Industry analysts echo these concerns. Gartner, which had once predicted sweeping AI-driven staff reductions, has recently reversed course, noting a trend toward rehiring human agents as organizations recognize the irreplaceable value of human empathy and adaptability.

As call centers continue to experiment with GenAI, the consensus is clear: while the technology can assist with routine queries and post-call tasks, its limitations in understanding context, emotion, and complex language mean that human oversight remains essential. The most effective strategy for balancing efficiency with quality customer care should  be a hybrid approach, where AI augments rather than replaces human agents… for now.